A useful UI prompt tells a generator what to make, who will use it, what the interface must contain, and how the team will judge the result. Treat it as a compact design brief—not a guarantee of a correct first draft. Separate required product behavior from optional visual direction, then inspect and refine the output against explicit criteria.
Start with the deliverable and the user’s task
Say whether you need a single screen, a multi-screen flow, a static mockup, or an interactive prototype. Identify the intended user and the task they need to complete. “Design a dashboard” leaves the audience, purpose, and expected result open to interpretation; “Create an interactive appointment-booking flow for patients choosing a clinic and time” gives the generator a defined target.
As an Amazon Associate I earn from qualifying purchases.
Figma’s guidance recommends direct, specific requests and illustrates how a concrete feature list can be more useful than a broad prompt. It also distinguishes prompting for a layout from adding functionality in later prompts. See Create a Figma Make file.
Turn requirements into things a reviewer can check
List the interface’s required content and controls, and make the hierarchy clear. Include exact visible text for labels that must not be paraphrased. For interactive work, specify important states such as loading, empty, error, and success; a control is easier to assess when its expected behavior is stated.
#1 Best Overall
- Screen or flow: Which page or sequence should appear?
- Content: What information, headings, and labels must be visible?
- Controls: What buttons, fields, menus, or navigation elements are required?
- Hierarchy: Which action or information should be most prominent?
- States: What should users see before, during, and after an action, including failure cases?
- Exclusions: What should not be added?
These details make evaluation more objective. OpenAI’s frontend guidance discusses tailoring an interface to its users and product domain while specifying controls and states. Its UI mockup evaluation example checks whether the requested screen and elements are present, the hierarchy is clear, text is readable, and controls communicate their affordances. Evaluate the output against the request, not requirements nobody supplied. See Frontend prompt instructions and Image Evals for Image Generation and Editing Use Cases.
Separate product requirements from visual direction
First make clear what the product must do and what the screen must contain. Then add optional styling guidance: for example, a calm visual tone, a particular density, or a reference to an existing design system. This distinction helps a team review whether the interface works as requested without mistaking stylistic preference for a functional requirement.
If the work must follow an established system, name it and provide applicable rules, components, or variables when the tool supports them. Figma documents guidelines for carrying design-system context into Figma Make, while cautioning that too much context can confuse rather than help. Include the rules relevant to this screen instead of pasting every available document. See Add guidelines to Figma Make.
Choose references that match the information you have
When using Figma Make, Figma recommends attaching a Figma frame rather than an image when possible: a frame can provide structured design data, while an image may communicate broad visual direction without reliably conveying exact details such as colors. Use a screenshot or image when it is what you have, but state which aspects matter—such as layout or mood—and specify exact values separately when they are essential. See Figma’s guidance on creating a Figma Make file.
For real-world content that needs to be current, distinguish that factual requirement from styling. Figma documents web search and URL fetching in supported Figma agent and Make workflows; those capabilities and their availability depend on the product workflow. If the prototype needs information from a page or public design system, identify the source to use rather than expecting a visual reference to provide verified facts. See Search the web with the Figma agent and Figma Make.
Use a prompt structure, not a magic formula
The following is an adaptable editorial template, not a vendor-prescribed formula. Remove irrelevant parts and fill in the specifics your team can review.
Rank #4
Create a [static mockup / interactive prototype] for [screen or flow] used by [audience] to [user task]. Include [required content and controls], in this order of importance: [hierarchy]. Follow [design system or reference] and use [visual direction]. Include these states: [loading, empty, error, success, or other relevant states]. Use the exact visible text for [critical labels]. Do not add [explicit exclusions]. Make the result reviewable by showing [screen/frame/deliverable].
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
Keep instructions task-specific and organized. OpenAI’s prompt-engineering guidance discusses separating general instructions from details and examples for the particular task; a concise brief with relevant constraints is more useful than an undifferentiated list. See Prompt engineering.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Pick a workflow that fits the deliverable
There is no universally best prompting approach. Choose based on what you need to review and what context the tool can use.
| Decision | Choose this when… | What to provide or check |
|---|---|---|
| Static mockup or functional prototype | You need either a visual composition or a UI with interactions reviewers can try. | State the deliverable explicitly. Figma describes Make for prototypes, web apps, and interactive UI; tool capabilities and access can change. Figma Make file guidance. |
| Structured frame or screenshot/image | You have an existing design file, or only a visual reference. | Prefer a Figma frame when available for its structured design data; use an image for broad visual cues and spell out exact details separately. Figma Make file guidance. |
| Existing system or greenfield styling | The result must match established components and rules, or the team is exploring a new direction. | Name the design system and relevant guidelines if consistency matters. For exploratory styling, describe the intended visual direction without implying that it is an existing standard. Figma guideline guidance. |
| Static knowledge or current, source-grounded content | The mockup uses generic example content, or it must reflect a real page or public design system. | For supported workflows, Figma documents web search and URL fetching. Identify the source and the content to use; verify current product support in its documentation. Figma web-search guidance. |
| One large prompt or staged prompts | The request is simple enough to specify at once, or its layout and behavior are easier to review separately. | For a complex interface, establish the layout first and then request functionality or focused changes. Figma recommends this staged approach in its Make guidance. Figma Make file guidance. |
Review the draft, then make focused changes
Generation is the start of review, not the end of design work. Check the result against the brief before adding more instructions. If it misses a requirement, ask for a targeted correction; if the layout is still unsettled, resolve that before layering in interactions. Figma also recommends adjusting generated results directly. See AI Design Prompts: How to Get Better Output.
- Is this the requested screen or flow, for the stated user task?
- Are all required elements present, and are excluded elements absent?
- Does the visual hierarchy reflect the stated priorities?
- Are the text and critical labels readable and exact?
- Can a reviewer tell what each control does?
- Does the draft follow the supplied design system or reference in the ways that matter?
- Are the requested interaction states represented?
Use the same checklist for each iteration. It makes disagreements concrete: reviewers can point to a missing control or unclear action instead of relying on a vague judgment that a mockup “feels off.”
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
